## page was renamed from CoP_chapter5 = 5. Interbull CoP - Methods of international evaluation = == 5.1. Multiple-trait Across Country Evaluation (MACE) == 5.1.1. International evaluations will be computed by a linear multiple-trait across country (MACE) model analysis of national evaluation results from the participating countries. . The model to be used is: '''Y = c + g + s + e''' . Where: Y = De-regressed national genetic evaluations c = Country of evaluation effect g = Genetic group of bull effect, defined by the bull’s population of origin and year of birth s = Bull genetic effect including genetic relationships among bulls in all participating countries e = Residual effect. The international predicted genetic merits will be formed by the sum of the solution for the bull, the genetic group and country effects. 5.1.2. International evaluations will be computed for one trait at a time, and within breed. For breed definition, see item 7.1.7. 5.1.3. Data on all bulls evaluated in each country will be considered in the international genetic evaluation, subject to the following editing criteria: a. only artificial insemination (AI) bulls are included; "other" bulls (status of bull = 20) are considered only if the predicted genetic merit is identified as based on second crop daughters only (type of genetic merit = 21) or imported daughters/embryos (type of genetic merit = 22; see file format for definitions of status of bull and type of predicted genetic merit), a. only bulls with daughters recorded in at least 10 herds are included, a. only bulls born since 1986 and 1981 for Holstein and other breeds, respectively, are included, a. bulls with second country evaluations (type of predicted genetic merit = 21) are included only when the predicted genetic merit meets national standards for official publication in the country of first evaluation of the animal (official publication of predicted genetic merit = Y and type of predicted genetic merit = 11 or 12) OR if the predicted genetic merit is based on minimum 150/30/80 daughters in 50/10/20 herds (Holstein/Guernsey/other breeds) whether or not first country evaluation (type of predicted genetic merit = 11 or 12) is included in the data. Second country data may be excluded in view of scientific evidence of bias, examined by the Interbull Centre a. bulls with second country evaluations based on more than 50% imported daughters/embryos (type of genetic merit = 22) are included only when the predicted genetic merit meets national standards for official publication in the country sending the information (official publication of predicted genetic merit = Y), and first country evaluation (type of predicted genetic merit = 11 or 12) is included in the data. Exceptions from these requirements may be accepted as part of Interbull standard procedures for a trait-group and specified in the appropriate section of Chapter 6 “Traits and breeds”. 5.1.4. Data on all bulls evaluated in each country will be considered in estimation of genetic correlations between countries, subject to the following editing criteria: a. only artificial insemination (AI) bulls are included; “other” bulls (status of bull = 20) are considered only if the predicted genetic merit is identified as based on second country evaluation (type of predicted genetic merit = 21; see file format for definitions of status of bull and type of predicted genetic merit). Exceptions from these requirements may be accepted as part of Interbull standard procedures for a trait-group and specified in the appropriate section of Chapter 6 “Traits and breeds”. 5.1.5. Dependent variables in the evaluation model are de-regressed national genetic evaluations. 5.1.6. Variation in the precision of the national predicted genetic merits is accounted for in the linear model by applying a weighting factor (effective daughter contribution, EDC) that considers amount of information, contemporary group structure, correlations of repeated observations within the same animal, and the reliability of daughters’ female ancestors. The calculation of EDC’s has to be implemented separately by each individual organization participating in Interbull evaluations, following the procedure in Appendix IV and V. 5.1.7. Within‑country sire variances are to be estimated during each evaluation run. Correlations between countries are estimated at every test run for all breeds and traits. Any change in national genetic evaluations should be tested in a test run as well as new countries/traits entering the system. 5.1.8. Conversion coefficients among all participating countries and for all traits are computed based on international predicted genetic merits of bull’s that are progeny tested only in one country (country of origin), with minimum birth year 1993 and 1992 for Holstein and other breeds, respectively in year 2004 (to be updated by 1 year at each January evaluation), and a predicted genetic merit based on a minimum 20 herds and with 75% international reliability/repeatability. A minimum of 20 bulls is required to compute such conversions. For country combinations not fulfilling these requirements conversion equations are computed based on theoretical b-values and empirical a-values. The theoretical b-value is computed as rG(i,j)×std(i)/std(j), where rG(i,j) is the genetic correlation between country i and j, and std(j) is the sire standard deviation for country j. The a-value is estimated as m(i)-b×m(j) where m(i) is the mean international proof for country i based on all bulls progeny tested only in one country born since 1996 (Holstein; 1995 for other breeds). Exceptions from these requirements may be accepted as part of Interbull standard procedures for a trait-group and specified in the appropriate section of Chapter 6 “Traits and breeds”. 5.1.9. Some details on the genetic evaluation procedure are provided in the service documentation made available at''' '''[[http://www.interbull.org/ib/maceev_archive|http://www.interbull.org]] and together with each distribution of results from the international evaluations 5.1.10. The international genetic evaluation procedure is based on international work described in the following scientific publications: a. International genetic evaluation computation: i. Schaeffer, 1994. J. Dairy Sci. 77:2671-2678 i. Klei, 1998. Interbull Bulletin 17:3-7 a. Weighting factors: i. Fikse and Banos, 2001. J. Dairy Sci. 84:1759-1767 a. De-regression: i. Sigurdsson and Banos, 1995. Acta Agric. Scand. 45:207-219 i. Jairath et al., 1998. J. Dairy Sci. Vol. 81:550-562 a. Genetic parameter estimation: i. Sigurdsson et al., 1996. Acta Agric. Scand. 46:129-136 i. Klei and Weigel, 1998. Interbull Bulletin 17: 8-14 i. Sullivan, 1999. Interbull Bulletin 22:146-148 i. Jorjani et al., 2003. J. Dairy Sci. 86:677-679 a. Time edits: i. Weigel and Banos, 1997. J. Dairy Sci. 80:3425-3430 5.11 a. International reliability estimation: i. Harris and Johnson, 1998. Interbull Bulletin 17:31-36 5.1.11. Newer developments regarding international evaluation methodology will be applied whenever research efforts are completed and considered appropriate, and approved by the Interbull Steering Committee. == 5.2. International Genomic Evaluation of Young Bulls (GMACE) == 5.2.1. International GMACE evaluations will be computed for one trait at a time, and within breed. For breed definition, see item 7.1.7. 5.2.2. Data on all bulls evaluated in each country will be considered in the GMACE international genetic evaluation, subject to the following editing criteria: a. National GEBV’s will be required to be from the same model and on the same base and scale as the national EBVs of progeny-tested bulls provided for classical MACE. A country can therefore only participate with genomic data in GMACE for the same traits as they participate with in classical MACE. 5.2.3. National GEBV data are edited according to the following criteria: . Bulls included in the GMACE breeding value predictions: * no conventional proof and maximum seven years of age * sire and dam with conventional proof in corresponding conventional evaluation or with GMACE evaluation (e.g. young genomic sire of young genomic bull) * pedigree in the Interbull pedigree data base * Genotypes data of bulls participating in international genomic evaluation are not subjected to edits that are based on the number of daughters/herds/EDC. 5.2.4. Dependent variables in the evaluation model are Mendelian Sampling deviations computed as '''MS = national GEBV – MACE PA'''. 5.2.5. Correlations between countries are estimated only in classical MACE but utilized also in GMACE. 5.2.6. The Parameter-space approach (Sullivan, 2016) is used in GMACE. Genomic variances are not estimated (the ratio of genomic SD to MACE SD is assumed to be equal to one). 5.2.7. Some details on the genetic evaluation procedure are provided in the service documentation made available at http://www.interbull.org/ib/gmace_archive and together with each distribution of results from the international evaluations 5.2.8. The GMACE procedure is based on international work described in the following scientific publications: a. GMACE implementation: i. Sullivan, P.G. 2016. Defining a Parameter Space for GMACE. Interbull Bulletin 50:85-93. i. Sullivan, P.G. and !VanRaden, P.M. 2010. Interbull Bulletin 41:3-7 i. Sullivan, P.G. et al., 2011. Interbull Bulletin 44: 87-94 i. Sullivan, P.G. and Jakobsen, J.H. 2012. Interbull Bulletin 45: 3-7. i. !VanRaden, P.M. and Sullivan, P.G. 2010. Gen. Sel. Evol. 42: 7 i. Sullivan, P.G. 2013. GMACE reliability approximation. Interbull Bulletin 47: 1-4 i. Sullivan, P.G. 2013. GMACE variance estimation. Interbull Bulletin 47: 5-9 i. Sullivan, P.G. 2013. GMACE weighting factors. Interbull Bulletin 47: 10-14. i. Sullivan, P.G. & Jacobsen, J.H. 2014. GMACE pilot #4: Adjusting the national reliability input data. Interbull Bulletin 48: 40-45. i. Sullivan, P.G. & Jacobsen, J.H. 2014. GMACE without variance estimation. Interbull Bulletin 48: 46-49. a. Validation of national genomic evaluations: i. Mäntysaari, E., Liu, Z and !VanRaden P. 2011. Interbull Bulletin 41, p. 17-21. == 5.3. International Genotype-based Evaluation (InterGenomics) == 5.3.1 International genomic evaluations will be computed by an iterative, nonlinear model with heavy-tailed prior for marker effects analogous to Bayes A (a curve parameter of 1.05 is used). 5.3.2 Base population allele frequencies are subtracted from genotypes, and a polygenic effect (poly) additive variance, which value is depending on the trait, is fit in the model. 5.3.3 The model applied is DPGM = mean + Σgenotypes*effects + poly + error. 5.3.4 The !InterGenomics procedure is based on international work described in the following scientific publications: * !VanRaden, P. (2008) Efficient methods to compute genomic predictions. Journal of dairy Science 91. 4114-4123 * !VanRaden, P. M. 2011. findhap.f90. Accessed May 25, 2011. http://aipl.arsusda.gov/software/findhap/